timm/qwen3_vit_306m.qwen3_vl_4b
Same weights, converted vs Qwen/Qwen3-VL-4B-Instruct. Same weights stored in another dtype (BF16 → F32), plus 97 new tensors.
Tensors changed: Attention 0/96, Norms 0/96, Embeddings 0/1
Same weights, converted vs Qwen/Qwen3-VL-4B-Instruct. Same weights stored in another dtype (BF16 → F32), plus 97 new tensors.
Tensors changed: Attention 0/96, Norms 0/96, Embeddings 0/1